guidelines for video delivery over a mobile network. for n excee speed stand such in vid fi limit of...
TRANSCRIPT
Copyright Ⓒ 2014 NTT DOCOMO, INC. All Rights Reserved.
Guidelines For Video Delivery Over A Mobile Network
--For efficient video delivery--
Version 1.0
July 1st, 2014
NTT DOCOMO, Inc.
Copyright Ⓒ 2014 NTT DOCOMO, INC. All Rights Reserved.
- 2 -
Contents
Chapter 1. Introduction .......................................................................................................... - 6 -
1.1. Target of this document .............................................................................................. - 6 -
1.2. Structure of this document ........................................................................................ - 6 -
1.3. References ..................................................................................................................... - 7 -
Chapter 2. Unique characteristics of mobile network services and actions taken by
DOCOMO - 8 -
2.1. Data volume in the mobile network by area and time period ............................ - 8 -
2.2. Limit of transmission speeds up to 128 kbps ........................................................ - 9 -
2.3. NTT DOCOMO’s actions to enhance its mobile network .................................. - 10 -
2.3.1. Quad-band LTE (efficient use of 4 spectrum bands) .................................. - 11 -
2.3.2. Installation of additional LTE base stations ............................................... - 12 -
Chapter 3. Image quality experienced by customers using video delivery service ................. - 13 -
3.1. Factors affecting the image quality of video delivery service .......................... - 13 -
3.2. Relationship between encoding conditions and image quality ........................ - 14 -
3.2.1. Relationship between encoding bitrate and image quality ...................... - 15 -
3.2.2. Relationship between image resolution/frame rate and image quality . - 17 -
3.3. Degradation in quality of experience in a congested mobile network ............ - 18 -
Chapter 4. Technological trends in video delivery .......................................................... - 20 -
4.1. High-compression codec technologies .................................................................... - 20 -
4.2. Video streaming technologies .................................................................................. - 23 -
4.2.1. HTTP streaming (pseudo streaming) ............................................................ - 24 -
4.2.2. HTTP adaptive streaming ............................................................................... - 25 -
Chapter 5. DOCOMO-Suggested measures for efficient network utilization ........... - 27 -
5.1. Optimization of content bitrate (image quality) ................................................. - 27 -
5.1.1. Optimization for H.264/AVC ........................................................................... - 27 -
5.1.2. Optimization for H.265/HEVC ........................................................................ - 28 -
5.2. Optimization of delivery bitrate (Quality of Experience) ................................. - 28 -
5.2.1. Video streaming technologies suitable for a mobile environment .......... - 28 -
5.2.2. Time zone-based delivery rate control .......................................................... - 30 -
5.2.3. UI (User Interface) enabling selection of delivery bitrate ........................ - 30 -
5.3. Other optimization measures .................................................................................. - 31 -
5.3.1. Adjustment of still images on web sites ........................................................ - 31 -
Chapter 6. Examples of measures taken for NTT DOCOMO’s video site .................. - 33 -
6.1. Delivery technologies and their usage tendency at “d-Amine Store” ............. - 33 -
6.1.1. Relationship between delivery technologies and image quality .............. - 33 -
6.1.2. DRM license validity period for Download scheme .................................... - 34 -
6.2. UI enabling selection of delivery bitrate .............................................................. - 34 -
Copyright Ⓒ 2014 NTT DOCOMO, INC. All Rights Reserved.
- 3 -
6.2.1. Video selection screen ....................................................................................... - 35 -
6.2.2. Image quality selection during playback ...................................................... - 35 -
6.3. Adjustment of still images on web sites ............................................................... - 35 -
6.3.1. Resizing for optimal resolution ....................................................................... - 35 -
6.3.2. Volume reduction (optimization) .................................................................... - 36 -
6.3.3. Utilization of browser cache ............................................................................ - 36 -
6.4. Notices for users ......................................................................................................... - 36 -
6.5. Video specifications at “d anime store” ................................................................. - 38 -
6.5.1. File specifications for HTTP streaming/Download technologies ............. - 38 -
6.5.2. File specifications for HLS technology .................................................................. - 40 -
6.6. 【Information】Introduction of tools ........................................................................ - 41 -
Copyright Ⓒ 2014 NTT DOCOMO, INC. All Rights Reserved.
- 4 -
About trademarks
· All the names of the companies, products, and services that are mentioned in this
document are trade names, trademarks, or registered trademarks of their respective
owners.
· This document does not specifically indicate any copyrights, or trademarks or
registered trademarks.
Copyright Ⓒ 2014 NTT DOCOMO, INC. All Rights Reserved.
- 5 -
Revision history
Version Section Type Revisions
1.0 ― First version
Copyright Ⓒ 2014 NTT DOCOMO, INC. All Rights Reserved.
- 6 -
Chapter 1. Introduction
This document describes ways of providing video content over a mobile network aimed
at providing comfortable video viewing experience for users. For comfortable video
viewing experience, it is important to enable users to watch high-resolution video that
runs smoothly without interruptions. However, the use of excessively high bitrates
compared with the forwarding bandwidth that the network can provide may cause video
images to skip or to be delayed, and this may result in deterioration in the service
quality experienced by users. Therefore, it is important, especially in a mobile network,
to maintain a suitable balance in the quality and size of images in video delivery.
While some measures are available to improve the quality of experience, the focus of
this document is placed on those that can be used to optimize content bit rate and
delivery bitrate. The first part of this document presents the fact that even for the same
video, its data size varies depending on the compression measures used, showing some
examples. The second part introduces some background information of a mobile
network and its varying traffic volumes according to the time of day, and then moves on
to discuss specific measures to control image quality based on the time of day. The third
part focuses on HTTP adaptive streaming technology, which can provide smooth and
seamless video streaming and efficiently utilize network resources in a mobile
environment.
The use of measures introduced in this document will help achieve efficient video
delivery. In addition to achieving comfortable video viewing experience to users, this
will generate circumferential benefits through the reduction of their traffic volumes in
the network. Efficient use of the finite resources in a mobile network will not only allow
users to watch video comfortably but also enable others around them to use the network
comfortably through suitable allocation of network resources. If your company is
currently providing or planning to provide video content, this document will serve as a
useful guide for achieving video delivery.
1.1. Target of this document
This document was written with video content providers in mind and, therefore, it
presumes that the reader has a basic knowledge of video delivery.
1.2. Structure of this document
Copyright Ⓒ 2014 NTT DOCOMO, INC. All Rights Reserved.
- 7 -
Chapter 2 of this document explains some unique characteristics of a mobile network.
Chapter 3 describes schemes for image quality evaluation. Chapter 4 explains
technological trends in video delivery. The background information provided in
Chapters 2, 3, and 4 will help readers understand the effectiveness of specific measures
introduced in Chapter 5. Those who are more interested in understanding specific
measures and implementation methods can directly start reading Chapter 5 before the
preceding Chapters 2, 3, and 4. Chapter 5 provides a systematic description of specific
measures. Lastly, Chapter 6 presents some measures actually implemented by NTT
DOCOMO as reference information. These examples are useful to have a better
understanding of the details described in this document.
Please note that this document does not cover descriptions of server applications used
for video delivery or terminal apps for video reception. This is due to an assumption that
server applications and terminal apps that can use the technologies described in this
document will be addressed by content providers engaged in video delivery.
1.3. References
[1] F. Dobrian, V. Sekar, A. Awan, I. Stoica, D. A. Joseph, A.Ganjam, J. Zhan, and H.
Zhang, "Understanding the impact of video quality on user engagement," Proc. of ACM
SIGCOMM'11, pp. 362-373, Aug.2011.
[2] ITU-T H.265 | ISO/IEC 23008-2 High Efficiency Video Coding
[3] ITU-T H.264 | ISO/IEC 14496-10 Advanced video coding for generic audiovisual
services
[4] Jun Okamoto and Takanori Hayashi, “Latest trends in image media quality
assessment technologies,” IEICE Fundamentals Review, Vol.6, No.4, pp.276-284, April
2013.(in Japanese)
[5] ITU-T Recommendation P.910, “Subjective video quality assessment methods for
multimedia applications,” April 2008.
[6] ISO/IEC 23009-1 Information technology – Dynamic adaptive streaming over
HTTP(DASH) – Part1: Media presentation description and segment formats
Copyright Ⓒ 2014 NTT DOCOMO, INC. All Rights Reserved.
- 8 -
Chapter 2. Unique characteristics of mobile network services
and actions taken by DOCOMO
2.1. Data volume in the mobile network by area and time period
Characteristics of a mobile network include varying peak hours depending on the
characteristics of areas* (see Figure 2-1 below). As of December 2013, the level of
congestion in the DOCOMO network peaks from 10 p.m. to 11 p.m. This is based on the
total volume of data generated by NTT DOCOMO’s Xi (LTE) service subscribers going
through the sp mode center. A closer analysis of smaller areas reveals that the peak
time varies in different areas. The peak time is from 12 noon to 1 p.m. in an area
covering a business district; from 7 a.m. to 8 a.m. and 6 p.m. to 7 p.m. at an urban
terminal station area; and from 12 midnight to 1 a.m. in a residential area. This shows
that each area experiences congestion in a different time zone. A failure to use an
optimal bitrate for video delivery in respective areas and time zones will increase the
risk of causing more intermissions and slowdowns in video. This makes users feel stress
during video viewing (lowering the quality of experience for users). Therefore, in order
to improve the quality of user experience, it is important to dynamically change the
delivery bitrate to optimize video delivery in accordance with the degree of congestion in
the network, using measures such as those hereinafter described.
*The “area” mentioned here is defined as an area covered by one or more base
stations.
2.2.
For N
excee
speed
stand
such
in vid
Fi
Limit of
NTT DOCO
eds a certai
d for the res
dpoint of en
limitation
deo delivery
gure 2-1 C
f transm
OMO’s Xi (L
in threshold
st of the mo
nsuring fair
are shown
y therefore
C
Changes in t
by
mission sp
LTE) service
d, a 128-kb
onth. This li
r network u
in Tables 2
may invite
Copyright Ⓒ
- 9 -
the data vol
area and ti
peeds up
e, if the volu
bps cap is im
imitation is
usage among
2-1 and 2-2.
the risk of
2014 NTT DOC
lume in a m
ime
to 128 k
ume of data
mposed on
one of the m
g users. Th
The use of
putting the
OMO, INC. Al
mobile netwo
bps
a used in th
the user’s u
measures de
e data volu
f an excessiv
user at a d
l Rights Res
ork
he current m
upload/dow
esigned from
umes that tr
vely high bi
disadvantag
served.
month
nload
m the
rigger
itrate
ge.
Copyright Ⓒ 2014 NTT DOCOMO, INC. All Rights Reserved.
- 10 -
Table 2-1 Summary of major Xi Packet Flat-rate services (currently available)
Xi Packet Flat-rate
Services
Max data volume for a
normal transmission speed
Requirement for
removing the 128kbps
cap
Xi Pake-hodai Light 3GB/month Payment of extra charges
for every 2GB data usageXi Pake-hodai Flat
7GB/month Xi Pake-hodai for iPhone
*For details of Xi Packet Flat-rate Servcies, please refer to NTT DOCOMO’s web page
(URL: https://www.nttdocomo.co.jp/charge/packet/index.html)
Table 2-1 Summary of major Packet Packs in new billing plans
(available from June 1, 2014)
Packet Pack Max data volume for a normal
transmission speed
Requirement for removing
the 128kbps cap
Raku-Raku Pack 200MB/month
Payment of extra charges
for every 1GB data usage
Data S Pack 2GB/month
Data M Pack 5GB/month
Share Pack 10 10GB/month (per group)
Share Pack 15 15GB/ month (per group)
Share Pack 20 20GB/ month (per group)
Share Pack 30 30GB/ month (per group)
*For details of the new billing plans “Kake-hodai & Pake-aeru,” which will help families
save money, please refer to NTT DOCOMO’s web page.
(URL:https://www.nttdocomo.co.jp/charge/new_plan/index.html?icid=CRP_TOP_main
PR_new_plan#charge)
2.3. NTT DOCOMO’s actions to enhance its mobile network
NTT DOCOMO is enhancing its mobile network by efficiently using four spectrum
bands (quad-band LTE) and installing more LTE base stations to support growing data
volume.
In addition to the actions above taken by NTT DOCOMO, content providers can also
play a role in making the users’ video watching experience much more comfortable by
using measures that will be described in Chapter 5 for video delivery. For actions taken
by NTT DOCOMO to enhance its mobile network, please refer to NTT DOCOMO’s web
page (URL: https://www.nttdocomo.co.jp/xi/index.html)
Copyright Ⓒ 2014 NTT DOCOMO, INC. All Rights Reserved.
- 11 -
2.3.1. Quad-band LTE (efficient use of 4 spectrum bands) NTT DOCOMO’s LTE “Xi” service efficiently utilizes the four spectrum bands - 1.7GHz,
1.5GHz, 800MHz, and 2GHz- in order to further enhance its mobile network.
The 2GHz and 800MHz bands have been used to provide “coverage” and the 1.5GHz
band to provide “speed.” In September 2013, the 1.7 GHz band was added to NTT
DOCOMO’s 3 operation bands mentioned above to launch its “quad-band LTE”
operation.
Figure 2-2 Image of quad-band LTE
Efficient use of 4 spectrum bands
In Tokyo, Nagoya, Osaka, more than double the spectrum In preparation 3G band LTE band
Platinum band
Platinum band
Before 19 September, 2013 20 September, 2013 April 2014
Copyright Ⓒ 2014 NTT DOCOMO, INC. All Rights Reserved.
- 12 -
2.3.2. Installation of additional LTE base stations NTT DOCOMO is installing more LTE base stations to expand its LTE “Xi” areas so
that connections will be available anytime, anywhere. The number of its LTE base
stations will be increased 1.7 times over the fiscal year 2014 (from approx. 55,300 at the
end of FY 2013 to approx. 95,300 at the end of FY 2014) and the number of 100Mpbs
base stations more than 10 times (from approx. 3,500 to 40,000 during the same period
as above) for further enhancement of its LTE network.
Figure 2-3 LTE Xi base station installation plan
LTE base stations to be more significantly increased
24,400
95,300
End of FY2013End of FY2012 End of FY2014 (target)
100Mbps
base stations
40,000 3,500
55,300
1.7 times increase of LTE base stations
10 times increase of 100Mpbs base
t ti
Copyright Ⓒ 2014 NTT DOCOMO, INC. All Rights Reserved.
- 13 -
Chapter 3. Image quality experienced by customers using
video delivery service
3.1. Factors affecting the image quality of video delivery service This section describes some factors that affect the quality of images experienced by
users (Quality of Experience) in video delivery service.
Figure 3-1 shows elements that constitute video delivery service using a mobile network
as well as major factors that affect the video image quality. Video data is encoded into a
format suitable for transmission over the network and stored in the delivery server of a
content provider. NTT DOCOMO’s network receives video data over the Internet, etc.
and forwards it via a radio link such as an LTE network to smartphones, tablets, and
other video viewing terminals. The terminal that receives the video data decodes the
data into images that are to be displayed on its screen.
The image quality of video delivery service is affected by different elements that
constitute video delivery service. Major factors that affect the image quality of video
delivery service are generally classified as follows:
(1) Image encoding conditions;
(2) Throughput characteristics of transmission of encoded data over a mobile network;
(3) Performance of the video viewing terminal.
First, (1) Image encoding conditions for video delivery include the following.
・Coding scheme
・Encoding bitrate
When the bitrate is set low, it results in lack of sharpness (blurring) or mosaic-like
distortions (, which is called “block-noise”) in images.
・Image resolution
When the image resolution is low, it reduces the definition of images.
・Frame rate
When the frame rate is low, it reduces the smoothness of motions, resulting in
flickering and jerky images.
Second, when (2) the throughput of the mobile network is too low, the download of video
data cannot catch up with the playback of video on the terminal. As a result, the
playback is halted until a certain amount of video data are stored again in the terminal
(, wh
loss,
Furth
quali
capab
make
term
The
insta
unne
conge
level
facto
Fig
3.2.
Unde
first
accor
cond
resol
hich is calle
packet tran
hermore, re
ity achieve
bilities. Co
es image de
minals are sh
factors affe
ance, a high
ecessarily h
estion (thro
l of congesti
ors.
gure 3-1 M
Relation
erstanding
and foremo
rdance with
itions herei
lutions, and
ed rebufferi
nsmission d
egarding (3
ed on the t
mpared wit
eterioration
howing vide
ecting imag
her encodin
high bitrate
oughput deg
ion in a mo
Major factors
nship bet
of the rela
ost importa
h the level o
inafter desc
d frame rat
C
ing). Such t
delay, and de
3) the perfor
terminal di
th smartph
n on tablet
eo with the
ge quality
ng bitrate g
e is used
gradation).
obile networ
s affecting t
service us
tween en
ationship be
ant to delive
of congestio
cribed inclu
tes to be ad
Copyright Ⓒ
- 14 -
throughput
elay variati
rmance of t
iffers depen
hones, table
terminals m
same image
are not m
generally im
for video d
Achieving e
rk requires
the element
sing a mobil
ncoding c
etween enco
er video to
on in a mob
ude the cod
dopted. In t
2014 NTT DOC
degradatio
ion that occ
the video vi
nding on t
et terminals
more easily
e quality,.
utually ind
mproves ima
data transf
efficient vid
a deep und
s and image
le network
condition
oding condi
users with
ile network
ing scheme
this section
OMO, INC. Al
on can be c
ur in the ne
iewing term
heir decodi
s have larg
y perceived
dependent b
age quality
fer, it may
deo delivery
derstanding
e quality of
ns and im
itions and i
optimized i
k. Please no
es, encoding
n, the relati
l Rights Res
aused by p
etwork.
minal, the i
ing and di
ger screens.
even when
but related
y. However,
y cause net
y considerin
g of such qu
f video deliv
mage qua
image qual
image qual
ote that enc
g bitrates, i
ionship bet
served.
packet
image
isplay
This
n both
d. For
if an
twork
ng the
uality
very
ality
lity is
ity in
oding
image
tween
encod
resol
3
Figu
The h
quali
of the
to th
secur
marg
ding bitra
lution/frame
3.2.1. Rela
re 3-2 show
horizontal a
ity. The hig
e original v
hat of the or
red. On the
ginal when
ate and i
e rate and i
ationship b
ws a genera
axis shows
gher the enc
ideo (the vi
iginal video
e other han
the encodin
C
image qua
image quali
between en
al relationsh
the encodin
coding bitra
deo before e
o can be obt
nd, the effe
ng bitrate is
Copyright Ⓒ
- 15 -
ality and
ity are discu
ncoding bi
hip betwee
ng bitrate w
ate is set, th
encoding). H
tained when
ect of imag
s set too hig
2014 NTT DOC
the rela
ussed.
itrate and
n encoding
while the ve
he closer the
However, th
n a certain l
e quality im
gh beyond n
OMO, INC. Al
ationship b
image qu
bitrate an
rtical axis s
e image qua
he image qua
level of enco
mprovemen
ecessity
l Rights Res
between i
uality
nd image qu
shows the i
ality gets to
ality compa
oding bitrat
nt becomes
served.
image
uality.
image
o that
arable
te are
more
The
schem
depe
when
defin
defin
bitra
motio
relat
1Thisdecodroom
Figure
value of en
mes. Even w
nding on en
n the same
nition and a
nition and a
ate. On the o
on require
tionships ar
s is becauseding and no
m for image
e 3-2 Relati
ncoding bitr
when the sa
ncoder impl
encoder is u
amount of m
a smaller a
other hand,
s a higher
re summariz
e the specificot encoding,quality imp
C
ionship betw
rate that giv
ame coding
lementation
used, the im
motion of t
mount of m
video with
r encoding
zed in Figur
cations (sta, trying to leprovement.
Copyright Ⓒ
- 16 -
ween encod
ves sufficie
scheme is
n (difference
mage qualit
the video. I
motion can a
a higher le
bitrate to
re 3-3.
andards) of veave encode
2014 NTT DOC
ding bitrate
nt image q
applied, the
e in the enc
ty is differen
n general,
achieve hig
vel of defini
o achieve h
video codiner manufact
OMO, INC. Al
and image
uality varie
e image qua
coding effici
nt dependin
video with
gh image qu
ition and a l
high image
g schemes oturers/softw
l Rights Res
quality
es among c
ality is diffe
iency). And
ng on the le
a lower lev
uality at a
larger amou
e quality. T
only target ware provid
served.
oding
erent1
d even
evel of
vel of
lower
unt of
These
ers
Copyright Ⓒ 2014 NTT DOCOMO, INC. All Rights Reserved.
- 17 -
Figure 3-3 Effects of implementation of coding schemes and characteristics of video on
the relationship between encoding bitrate and image quality
3.2.2. Relationship between image resolution/frame rate and image quality
An increasing number of content providers are handling high-resolution images such as
those for high-definition television (HDTV) in order to provide users with
high-resolution video delivery service. The current mobile network environment,
however, is not always fully prepared to ensure the bandwidth sufficient for delivery of
HDTV-class video. This makes necessary to select a suitable image resolution and frame
rate in order to achieve higher video quality with a smaller bandwidth. Coding with a
low image resolution can yield high quality image in a low bandwidth. This relationship
is shown in Figure 3-4. In an example illustrated in Figure 3-4, the encoder using the
bitrates B1, B2, and B3 achieves their optimal video image quality for the image
resolutions of 320×240, 640×480, and 1280×720, respectively. When an encoding bitrate
lower than a level of 100 to 200kbps is used, reducing the frame rate is effective from
the standpoint of maintaining the image quality. However, reducing the frame rate in
fast-motion video such as of a sport scene will make the flickering and jerky parts more
noticeable. Degradation in the quality of experience can be minimized for content like
video telephony even at a low frame rate. However, for content of ordinary video
delivery service, a frame rate of approximately 15 frames per second is at least
necessary.
3.3.
This
occur
Figure
Degradnetwork
section des
rs when
3-4 Relati
ation in k
scribes char
a mob
C
ionship betw
quality
racteristics
bile netwo
Copyright Ⓒ
- 18 -
ween image
of exper
of degrada
ork is
2014 NTT DOC
e resolution
rience in
ation in the
congested.
OMO, INC. Al
and image
n a conge
e quality of
As wa
l Rights Res
quality
ested mo
f experience
as shown
served.
obile
e that
n in
Figu
imag
impr
quali
in a
deter
whic
kbps
cause
not o
watc
etc.).
the q
re 3-2 abov
ge quality
rovement co
ity is achiev
congested
rioration in
h degradati
s. In both ca
es video pla
only degrad
ching time (v
. Therefore,
quality of vi
Figur
ve, when a
gradually
omes to satu
ved with a h
mobile net
the user’s q
ion in the q
ases, the us
ayback to st
des the qual
video viewin
it is necess
ideo images
re 3-5 Degr
C
sufficient
improves
uration at a
higher enco
work due t
quality of ex
uality of exp
se of encodi
tall, which
lity of exper
ng behavior
sary to prev
.
radation in
Copyright Ⓒ
- 19 -
throughput
as the en
a certain po
oding bitrat
to insufficie
xperience. F
perience oc
ing bitrate
lowers the
rience but a
r) of users a
vent video s
quality of e
2014 NTT DOC
t is secured
ncoding bit
int. Howeve
te, video pla
ent through
Figure 3-5 d
curs at the
larger than
quality of e
also greatly
according to
stall in orde
experience d
OMO, INC. Al
d in a mobi
rate is rai
er, even wh
ayback may
hput. This c
demonstrate
throughput
n the respec
experience.
y affects the
some repor
er to maint
due to video
l Rights Res
ile network
ised until
hen higher i
y frequently
causes a dr
es an examp
ts of 300 an
ctive throug
Such video
e length of
rts (referenc
tain and im
o stall
served.
k, the
such
image
y stall
rastic
ple in
d 500
ghput
o stall
video
ce [1],
prove
Copyright Ⓒ 2014 NTT DOCOMO, INC. All Rights Reserved.
- 20 -
Chapter 4. Technological trends in video delivery
4.1. High-compression codec technologies
The latest video compression standard HEVC or officially called “ITU-T H.265 |
ISO/IEC 23008-2 High Efficiency Video Coding[2]” was approved by the ITU in January
2013. HEVC is an all-purpose standard designed to address a wide range of applications
from video on small screens of mobile devices to high-definition video with HD
resolution of 4K, 8K or higher. The HEVC standard was developed to double the
compression rate compared with the conventional H.264/AVC standard[3].
First, this section describes the terms “Profile,” “Level,” and “Tier,” which denote the
playback capability of an H.265/HEVC decoder.
The current H.265/HEVC standard defines the following three Profiles:
-Main Profile;
-Main 10 Profile;
-Main Still Picture Profile.
Both of Main Profile and Main 10 Profile are profiles for video compression, using
different bit depths to express one pixel. Main Profile only supports video with a depth
of 8 bits per pixel while Main 10 Profile also supports 10 bits per pixel in addition to 8
bits per pixel. Except for bit depth coding, the both Profiles use common technical
elements for coding. Both of them only support the chroma subsampling format of 4:2:0
(in which the amount of information on color difference is a quarter of that of
brightness), which is for consumer applications. An extension of the H.265/HEVC
standard called RExt is being standardized to support formats with a large color
difference, such as 4:2:2 and 4:4:4, which are mainly used for material transmission in
broadcasting stations. HEVC RExt also aims to support video with a bit depth higher
than 10bits per pixel. Thus, HEVC RExt is a standard mainly targeted at compression
and transmission of high-definition video, such as digital cinema and material
transmission. As for interlaced video, Main Profile and Main 10 Profile can handle this
type of video . However they haven’t introduced any coding tools specialized for
interlaced video and currently only use flags to mark it in the stream. In actual
processing, interlaced video is encoded as progressive video with its top and bottom
fields separated.
Next part discusses the Level and Tier. There are 13 Levels from 1 to 6.2, each of which
mainly specifies the maximum resolution and frame rate. The Tier, on the other hand, is
Copyright Ⓒ 2014 NTT DOCOMO, INC. All Rights Reserved.
- 21 -
divided into two, the main Tier and the high Tier, each of which specifies the maximum
bitrate at each level. H.264/AVC does not have a concept of Tier but specifies
everything including the maximum resolution, frame rate, and bitrate, using the Level
only. However, this makes it difficult to address a case where only the bitrate should be
restricted. H.265/HEVC, which has newly introduced the concept of Tier, is able to
specify the restriction on the resolution, frame rate, and bitrate, separately. Table 4-1
below shows a list of major resolutions, and Levels and Tiers.
Table 4-1 List of main resolutions, Levels, and Tiers
Main
Levels
Examples of maximum
resolution and frame rate
Maximum video bitrate [kbps]
Main Tier High Tier
1 QCIF 15fps 128 -
2 CIF 30fps 1,500 -
3 QHD(960x540) 30fps 6,000 -
4 2Kx1080(2048x1080)
30fps
12,000 30,000
5 4096x2160 30fps 25,000 100,000
6 8192x4096 30fps 60,000 240,000
6.2 8192x4096 120fps 240,000 800,000
Listed below are specific examples of coding technologies used in H.265/HEVC.
- Block partitioning
- Inter and intra prediction
- Orthogonal transform
- Variable-length coding
- In-loop filter
As is the case with major conventional coding standards, H.265/HEVC adopts a
compression scheme based on a combination of motion compensating prediction and
orthogonal transform. H.265/HEVC doubles the compression performance of H.264/AVC
by accumulating improvements in individual coding tools. However, H.265/HEVC has
no compatibility with conventional standards. Furthermore, its complexity gradually
adds more processing load, especially at the time of compression. To cope with this,
researchers and developers are working on to develop an H.265/HEVC encoder capable
of high-speed operation.
As shown in Figure 4-1, a hierarchical block portioning structure (Coding Tree Unit;
CTU) is introduced to H.265/HEVC to enable efficient coding of video images in widely
varying levels of resolution. Each block in the CTU is further split into subdivisions
called PUs (Prediction Units). By combining all these blocks, a wide variety of blocks
can b
64x6
The
parti
leadi
As fo
drast
comb
predi
meas
vecto
rates
A hie
to co
comp
As fo
Arith
used
schem
Ca
be used for
64.
introductio
itioning an
ing to comp
or intra pre
tically incre
bination wit
iction), whi
sures. They
or candidat
s.
erarchical s
ombine four
pression effi
or variable-l
hmetic Codi
d the Conte
me, in spite
Com
par
ategorization b
A
coding, ran
Figure 4-1
on of this m
d optimized
ression per
ediction, the
eased to 35
th the block
ch has also
y include the
es. Those m
structure ha
r sizes of ort
iciency.
length codin
ing (CABAC
ext-based A
e of its low
mplex block
rtitioning
Symmetrical
by PU
Asymmetrical
C
nging from
1 Block par
method has
d coding ac
formance im
e number of
from 9 for H
ks of differe
increased b
e use of a n
measures h
as also been
thogonal tr
ng, the only
C), which ha
Adaptive Va
w compressio
l partitioning
l partitioning
Copyright Ⓒ
- 22 -
the minimu
rtitioning m
enabled H
ccording to
mprovemen
f prediction
H.264/AVC.
ent sizes des
block size op
newly create
ave contrib
n introduce
ransform fro
y scheme av
as high com
Variable Len
on efficienc
g
g
2014 NTT DOC
um size of 4
method of H.
H.265/HEVC
o image cha
nt.
n modes ava
This allow
scribed abov
ptions, has
ed merge m
buted to ach
ed to orthog
om 4x4 to 3
vailable is th
mpression pe
ngth Codin
cy. On the o
OMO, INC. Al
4x4 to the m
265/HEVC
C to perform
aracteristics
ailable for H
s more prec
ve. Inter pr
introduced
ode and mu
hieving high
gonal transf
32x32 has g
he Context-A
erformance
ng (CAVLC)
other hand,
l Rights Res
maximum s
m complex
s in each b
H.265/HEVC
cise predicti
rediction (m
some innov
ultiple pred
gher compre
form. The a
greatly imp
Adaptive B
e. H.264/AVC
), a lightw
, this option
Hierarchy 0
Hierarchy 1
Hierarchy 2
Hierarchy 3
served.
size of
block
block,
C has
ion in
motion
vative
iction
ession
ability
roved
Binary
C has
weight
n has
Copyright Ⓒ 2014 NTT DOCOMO, INC. All Rights Reserved.
- 23 -
been excluded and is not available for H.265/HEVC in consideration of the progress of
device performance in recent years.
For an in-loop filter, the deblocking filter for block noise reduction and the Sample
Adaptive Offset (SAO) for ringing and pixel value shift control have been introduced.
These have contributed to achieving a reduction of coding distortion.
Table 4-2 provides a comparative summary of coding tools between H.265/HEVC and
H.264/AVC.
Table 4-2 Comparative summary of coding tools of H.264/AVC and H.265/HEVC
Coding tool H.264/AVC (High Profile) H.265/HEVC
Intra prediction 4x4 block: 9 modes
8x8 block: 9 modes
16x16 block: 4 modes
35 modes for 4x4 to 64x64
block sizes
Inter prediction (motion
prediction)
4x4 to 16x16 blocks
Quarter-pel accuracy
search
8x4/4x8 to 64x64 blocks
Quarter-pel accuracy
search
Orthogonal transform 4x4 or 8x8
Integer accuracy DCT
4x4 to 32x32 (4 types)
Integer accuracy DCT
Variable coding 2 types:
・ Context-based Adaptive
Variable Length Coding
(CAVLC)
・ Context-Adaptive Binary
Arithmetic Coding
(CABAC)
1 type:
・ Context-Adaptive Binary
Arithmetic Coding
(CABAC)
In-loop filter 1 type:
・Deblocking filter
2 types:
・Deblocking filter
・ Sample Adaptive Offset
(SAO)
4.2. Video streaming technologies
Video streaming technologies are available for delivery of video content over a network .
Video streaming technologies are mainly divided into two groups depending on the
protocols they use: HTTP and non-HTTP (RTSP:Real-Time Streaming Protocol, etc.).
Video delivery services today use HTTP-based streaming technologies. This is because
Copyright Ⓒ 2014 NTT DOCOMO, INC. All Rights Reserved.
- 24 -
the HTTP protocol has excellent conformity with standard web servers and existing
server systems in that it can make use of HTTP cache. HTTP is also highly compatible
with user systems in that it can transparently go through firewalls and NATs. Most
well-known HTTP-based streaming technologies are HTTP streaming and HTTP
adaptive streaming.
In video delivery, the quality of experience of users is affected by the quality of video
(resolution, frame rate, etc.) and the smoothness of playback. Raising the quality of
video increases the amount of data (bitrate) flowing over the network. The more the
video bitrate exceeds the transfer speed, the more frequently video playback will be
interrupted. In this manner, video quality and playback smoothness are in the
relationship of trade-off. Especially in a mobile environment, characterized by unstable
and highly fluctuating transfer speed, our challenge is how to ensure playback
smoothness. HTTP adaptive streaming is aimed at mitigating such trade-off.
4.2.1. HTTP streaming (pseudo streaming) This is the most widely utilized streaming technology in video delivery service. In HTTP
streaming, when the user starts video playback, it triggers the download of a selected
video file from the web server. A video playback application (web browser or media
browser) performs sequential playback of the buffered video data in parallel with the
download of the video file.
The bitrate of the downloaded video data is fixed from the start to the end of its
playback. When the bitrate of the specified video data is high, it takes a long time until
the start of playback due to a large amount of data to be buffered. Although the amount
of data in the buffer is reduced as the playback proceeds, the playback will be
interrupted if the buffer is exhausted. To restart the playback, it is necessary to obtain
buffered data once again. If the user cancels the playback at this point, the data not yet
played back will be discarded. This means that the resources of the server and network
used for the download will be wasted.
4
To d
using
techn
envir
Even
strea
deliv
techn
make
befor
Furth
time
This
Ther
4.2.2. HTTdeliver high
g various n
nology tha
ronment ha
n in a mob
aming techn
vering vide
nology can
es it possib
re the start
hermore, in
, but send a
reduces th
refore, it can
Buffering
in a poor i
TP adaptivh-quality vi
etwork env
at enables
as been deve
bile networ
nology can
o data wit
reduce the
ble to reduc
of playback
n this techn
a part much
he amount o
n effectively
The amo
smooth p
may be reduc
internet envir
C
Figure 4
ve streamideo to a v
ironments (
switching
eloped. The
rk, which f
achieve al
th a bitrat
gap betwe
ce the amou
k.
nology, the
h data of the
of data disc
y reduce wa
ount of buffe
layback
ced to this lev
ronment and
Copyright Ⓒ
- 25 -
4-2 HTTP st
ing variety of d
(Wi-Fi, 3G/
of bitrat
technology
features ev
most unint
te suitable
een transfer
unt of buffe
server nee
e file as requ
carded even
asted of the
ering necessa
vel
2014 NTT DOC
treaming
devices (PC
LTE, fixed
tes accordi
y is called H
ver-changing
terrupted, s
e to the n
r speeds an
er and thus
d not send
uested by th
n when the
server and
ary for
OMO, INC. Al
s, TVs, sm
networks, e
ng to s
HTTP adapti
g network
smooth vide
etwork env
nd video dat
s shorten th
an entire v
he device rec
user cance
network.
l Rights Res
martphones,
etc), a strea
specific net
ive streami
conditions,
eo streamin
vironment.
ta bitrates.
he time req
video file a
ceiving the
els the play
served.
etc.)
aming
twork
ng.
, this
ng by
This
This
quired
at one
video.
yback.
C
Figure 4-3 H
Copyright Ⓒ
- 26 -
HTTP adapti
Do
2014 NTT DOC
ive streaming
ownload only
OMO, INC. Al
g
as much as n
l Rights Res
necessary
served.
Copyright Ⓒ 2014 NTT DOCOMO, INC. All Rights Reserved.
- 27 -
Chapter 5. DOCOMO-Suggested measures for efficient
network utilization
This chapter introduces some recommended measures that should be considered for
adoption to achieve efficient use of a mobile network for video delivery.
5.1. Optimization of content bitrate (image quality)
5.1.1. Optimization for H.264/AVC Table 5-1 shows a general relationship between the video image quality and the
DOCOMO-suggested image resolution and encoding bitrate for the H.264/AVC coding
technology. The “image quality” in the table means the quality of images experienced by
the user (quality of experience) who watches the video. The image quality in the table is
quantified based on the subjective perception (subjective quality) of humans users, such
as “clear image,” “smooth video,” and “image quality degradation perceptible but not
annoying.”2 Table 5-1 shows a result of a case using Main Profile as the profile, a frame
rate of 30fps, and a random access interval of approximately one second (32 frames). As
described earlier, when the encoding bitrate becomes low, the use of low image
resolution helps reduce image quality degradation. The table shows a sample case in
which the image resolutions are changed corresponding to the encoding bitrates.
Table 5-1 Relationship between image quality and DOCOMO-suggested image
resolution/encoding bitrate (in case of H.264/AVC)
Coding scheme:
H.264/AVC Image quality
Low Middle High
Suggested image
resolution
360p-equivalent
(640×360)
360p-equivalent
(640×360)
720p-equivalent
(1280×720)
Suggested video
encoding bitrate
S: approx. 350kbps
T/P: approx. 400kbps
S: approx.700kbps
T/P: approx. 800kbps
S: approx. 1.4Mbps
T/P: approx. 1.6Mbps
2The image quality was determined based on the results of evaluation conducted using the ACR Method (Absolute Category Rating Method), which is an evaluation method adopted as an international standard. Specifically, the quality of images of different types of content were evaluated. The suggested conditions for High, Middle, and Low image quality levels are derived from the content which showed average quality characteristics and evaluated to have ratings of 4, 3, and 2 respectively on the MOS (Mean Opinion Score) scale. For evaluation methods quantifying subjective image quality, please refer to the references [4] and [5], etc.
Copyright Ⓒ 2014 NTT DOCOMO, INC. All Rights Reserved.
- 28 -
* S stands for Smartphones: evaluation results from video viewing on 4.7-inch smartphones.
T/P stands for Tablets and PCs: evaluation results from video viewing on 15.5-inch notebook PCs.
* The values of the suggested video encoding bitrates are subject to change.
5.1.2. Optimization for H.265/HEVC
An effective measure to achieve high image quality at a lower encoding bitrate than
that of H.264/AVC is to use a coding scheme with a high compression capability. The
latest H.265/HEVC coding scheme discussed in Chapter 4 is said to almost double the
compression efficiency compared to H.264/AVC. Therefore, H.265/HEVC is expected to
be widely utilized in the mobile video delivery market in the future. Table 5-2 shows a
general relationship between image quality and suggested image resolutions and
encoding bitrates for the current H.265/HEVC coding technology. The table shows a
result of study of a case that uses Main Profile as the profile, a frame rate of 30fps, and
a random access interval of approximately one second (32 frames). Future
improvements in H.265/HEVC implementation technology are expected to achieve even
higher compression levels.
Table 5-2 Relationship between image quality and DOCOMO-Suggested image
resolution/encoding bitrate (in case of H.265)
Coding scheme:
H.265/HEVC Image quality
Low Middle High
Suggested image
resolution
360p-equivalent
(640×360)
360p-equivalent
(640×360)
720p-equivalent
(1280×720)
Suggested video
encoding bitrate
S: approx. 200kbps
T/P: approx. 200kbps
S: approx.350kbps
T/P: approx. 400kbps
S: approx. 700kbps
T/P: approx. 800kbps
* S stands for Smartphones: evaluation results from video viewing on 4.7-inch smartphones.
T/P stands for Tablets and PCs: evaluation results from video viewing on 15.5-inch notebook PCs.
* The values of the suggested video encoding bitrates are subject to change.
* In H.265/HEVC, when a high image resolution is used, coding may take a long time.
5.2. Optimization of delivery bitrate (Quality of Experience)
5.2.1. Video streaming technologies suitable for a mobile environment Table 5-3 shows the status of support on Android and iOS platforms for the streaming
technologies discussed in Section 4.2. HTTP Adaptive Streaming has multiple
technologies. Although they share the same basic concept of switching content files
according to the state of network, they adopt different file formats. MPEG-DASH
(Dynamic Adaptive Streaming over HTTP)[6] is an ISO/IEC standard and is officially
supported on Android4.4 and higher. HLS (HTTP Live Streaming) is a proprietary
Copyright Ⓒ 2014 NTT DOCOMO, INC. All Rights Reserved.
- 29 -
technology of Apple Inc., although it is supported on both Android and iOS platforms.
These two HTTP Adaptive Streaming technologies and HTTP streaming are the three
technology options for video delivery for mobile devices. (As stated in Section 4.2, in
light of a mobile environment, the adoption of HTTP Adaptive Streaming (MPEG-DASH
or HLS) is expected to improve user experience.)
This section describes DOCOMO-suggested operation methods for each of the HTTP
Streaming and HTTP Adaptive Streaming technologies in a mobile environment.
Table 5-1 HTTP-based streaming technologies
Technology Specified
By
Support on Android
(*1)
Support on
iOS
HTTP streaming - Supported Supported
HTTP
adaptive
streaming
MPEG-DASH ISO/IEC Android 4.4 and
higher (*2)
Not supported
HLS Apple Android 4.0 and
higher
Supported
Smooth Streaming Microsoft Not supported Not supported
HTTP Dynamic
Streaming
Adobe Not supported Not supported
(*1) The support status is based on a standard Android implementation. For the support
status on smartphones, please refer to products’ specifications.
(*2) Android 4.4 provides a video playback API, which will be necessary to support
MPEG-DASH. The use of MPEG-DASH in video delivery requires implementation of
MPD (Media Presentation Description) file analysis and processing of adaptive
streaming logic using the above-mentioned video playback API on the video playback
application.
5.2.1.1. HTTP adaptive streaming
In the ever-changing state of the mobile communication environment, it is suggested
that HTTP Adaptive Streaming technology be adopted for its ability to provide smooth
and uninterrupted video streaming and effectively utilize network resources. While
HLS and MPEG-DASH are supported on Android OS and iOS, it is requested that
which technology to choose be determined based on whether it is supported by the
target Operating System(s) in video delivery, the content generation tools used, and the
delivery server.
To take advantage of the benefits of Adaptive Streaming technology, it is suggested that
a set of three video stream files in three different bitrates be prepared (in case of HLS,
Copyright Ⓒ 2014 NTT DOCOMO, INC. All Rights Reserved.
- 30 -
four files(bitrates) because a file of an audio-only stream is also included). While the
relationship between image resolution and bitrate was explained in Section 5.1., it is
requested that the image resolution level in each stream be adjusted with some margins
added to lower and higher rates to make a wider range of bitrate available. As a
reference, the bitrates used in “d anime store” are shown in Section 6.5.
(Note) DOCOMO’s 2014 summer models do not support HTTP Adaptive Streaming
using video data encoded with H.265/HEVC. Therefore, it is requested that any data
using HTTP Adaptive Streaming be encoded with H.264/AVC.
5.2.1.2. HTTP streaming (pseudo streaming)
For video delivery using HTTP streaming technology, it is suggested that content be
made available at several different bitrates and image resolutions to allow users to
choose the image quality they prefer according to the states of their network and usage.
Please refer to Section 5.1 for the relationship between image resolution and bitrate.
Please refer to Section 6.5 for the operation parameters used in DOCOMO’s d anime
store.
5.2.1.3. Download scheme
The Download scheme is highly effective to enable users to comfortably watch video in a
mobile environment. Users can experience stress-free playback when they are on the
road and in an unstable network environment by downloading video data in a stable
network environment in advance.
5.2.2. Time zone-based delivery rate control In Chapter 3, a drastic degradation of image quality resulting from video stall was
explained. This problem occurs in a congested mobile network even when the encoding
bitrate is raised to improve image quality. In order to avoid such degradation, it will be
effective to perform the following delivery rate control after preparing data sets of the
same video in multiple encoding bitrates. In doing so, please use the peak time zones in
Section 2.1 and suggested encoding bitrates in Section 5.1 as reference.
1. Choose video data encoded in a low bitrate for a time zone where a mobile network
is congested
2. Do not use an unnecessarily high coding bitrate for access from a mobile network
even in a time zone where the mobile network is not congested.
5.2.3. UI (User Interface) enabling selection of delivery bitrate To enable users to smoothly change bitrate according to their network state even in the
middle of playback is one of the simple and effective means to improve quality of
Copyright Ⓒ 2014 NTT DOCOMO, INC. All Rights Reserved.
- 31 -
experience of users. This approach mitigates users’ dissatisfaction because they are
given power to make a decision during video viewing as to which to prioritize, image
quality or smooth playback.
The above-mentioned measure is suggested to be used in fixed –rate video delivery such
as HTTP streaming and progressive download.
For an example of how to operate delivery bitrate selection and switching functions at d
anime store , please refer to Section 6.2.
5.3. Other optimization measures
This section describes other suggested measures not covered in the previous sections.
5.3.1. Adjustment of still images on web sites In video delivery service, still images on the web sites to introduce video have the
second-largest volume of data following video images themselves. Although the use of
rich still images is effective in introducing video, slow browsing caused by “heavy” web
sites is a serious issue that can prompt users to cancel services. For the use of still
images on web sites, the following are four generally-known approaches to be taken. It
is strongly requested to consider their introduction in combination with your video
delivery schemes to improve quality of experience perceived by users visiting web sites.
5.3.1.1. Resizing for optimal resolution One still image is used in various sizes in web sites. When only the visual quality of
image is taken into consideration, generating a reduced version of a big file will suffice.
From the standpoint of disk capacity, however, maintaining multiple sizes of one still
image may not be always preferred. On the other hand, sending an image in an
appropriate size greatly reduces the volume of data. Data volume reduction leads not
only to shorter transfer time but also to faster and smoother drawing processing on
smartphones. This improves the quality of experience of users.
5.3.1.2. Image volume reduction (optimization) One of the effective measures well known since the feature-phone age is the utilization
of image volume reduction (optimization) tools. These tools can reduce the volume of
images by modifying them to an extent difficult to recognize with human eyes. Data
compression methods used by tools include increasing compression ratio in JPEG, use of
PNG24 or PNG8 in PNG, and deletion of header information such as Exif information.
Copyright Ⓒ 2014 NTT DOCOMO, INC. All Rights Reserved.
- 32 -
5.3.1.3. Utilization of browser cache Utilizing browser cache, known as a measure for faster web browsing, is effective for
images and large-sized JavaScript/CSS files.
Setting cache control of the browser on a web server used for delivery of images enables
the same images to be displayed when users open the web page again. This is done
without the need for transfer (or with a request with an if-modified-since header) or
download of images.
5.3.1.4. Delayed image loading
In web pages, images are used even outside the display area. A mechanism to delay
loading of those images until the time they enter the display area can shorten the
loading time required to complete a screen transition.
There are some plugins publicly available (e.g., Lazy Load Plugin for jQuery), which can
be used to facilitate the introduction of the above-mentioned mechanism and to improve
the quality of the experience for users.
Copyright Ⓒ 2014 NTT DOCOMO, INC. All Rights Reserved.
- 33 -
Chapter 6. Examples of measures taken for NTT
DOCOMO’s video site
This chapter provides specific examples of how the measures such as those described in
Chapter 5 have been implemented by NTT DOCOMO to utilize its network efficiently.
6.1. Delivery technologies and their usage tendency at “d-Amine Store”
In accordance with the description in Section 5.2, “d anime store” (DOCOMO’s anime
store) offers two video delivery formats: one for Android and the other for iOS. HTTP
streaming and Download are offered for Android and HLS and Download for iOS. In
addition, the Store offers three image quality levels.
6.1.1. Relationship between delivery technologies and image quality In general, anime users ask for high quality images. At d anime store, however, some
different tendency has been observed among its users. Table 6-1 shows usage ratios for
different image quality levels at d anime store. According to Table 6-1, in case of 3G/LTE,
more users choose the image quality that ensures stable playback. In case of Wi-Fi,
which secures sufficient throughput, they give greater importance to higher image
quality. The data shown in this table indicate that users choose the delivery scheme and
image quality the most suitable to their environment. It also shows that they are
actively utilizing Wi-Fi to view high-quality video.
Copyright Ⓒ 2014 NTT DOCOMO, INC. All Rights Reserved.
- 34 -
Table 6-1 Ratio of number of playback times for each usage type
(d anime store, second half of 2013)
Ratio of number of playback
times
Image quality
Total Normal Good Excellent
HTTP
streaming
3G/LTE Approx.
30%
Approx.
20%
Less than
10%
Approx.
40%
Wi-Fi Less than
10%
Approx.
10%
Approx.
20%
Approx.
30%
Download
3G/LTE Approx.
10%
Less than
10%
Less than
10%
Approx.
10%
Wi-Fi Less than
10%
Less than
10%
Approx.
10%
Approx.
10%
Total Approx.
40%
Approx.
30%
Approx.
30% 100%
(Note) Totals may not add up due to rounding.
6.1.2. DRM license validity period for Download scheme For Download, d anime store issues a DRM license valid for 48 hours from the time of
download. This service design enables the user to play video downloaded on a previous
day without generating any further communication. In addition, even after 48 hours
from the download, the user can play the downloaded video only by reacquiring their
DRM license file. This makes the user feel secure in using the service even when their
network environment cannot offer sufficient throughput for video delivery.
6.2. UI enabling selection of delivery bitrate
At d anime store, the user is able to select the delivery bitrate on two screens as shown
in Figure 6-1.
6
The u
time
to se
At th
playb
6
Whil
displ
differ
playb
start
bitra
6.3.
6
“d an
video
This
1.
Figure
6.2.1. Videuser can sel
of playback
lect any im
he same tim
back button
6.2.2. Imale on the v
lays a pop-u
rent deliver
back. The in
t playing fr
ate.
Adjustm
6.3.1. Resinime store”
o. When one
enables del
Video selec
(“d-Mark
(Introducto
image)
e 6-1 UI e
eo selectionlect a level o
k. There is
age quality
me, a link t
ns in order t
ge qualityvideo playb
up screen fo
ry bitrate a
nterworkin
om the last
ment of s
izing for op” uses CMS
e image is r
livery of da
ction screen
ket”)
ory
C
nabling sele
n screen of image qu
no default s
at one touc
to descripti
to keep the
y selection back screen
or image qu
according t
g with the
t point the
till imag
ptimal resS(Contents M
registered, t
ta in an opt
2. Video
(
Copyright Ⓒ
- 35 -
ection of de
uality from “
setting. The
ch of a butto
ion of each
user aware
during pln, pressing
uality selecti
to the chan
resume (bo
user was w
ges on we
solution Manageme
the system
timum size
o playback
(Media Play
(Video)
2014 NTT DOC
livery bitra
“normal,” “
e screen is
on.
image qua
e of image qu
ayback g the imag
ion. This al
nging netwo
ookmark) fu
watching be
eb sites
nt System)
generates f
for each scr
screen
yer)
OMO, INC. Al
ate (d anime
good,” or “
designed to
lity is place
uality.
e quality s
lows the us
ork conditio
unction enab
efore the ch
to manage
files in 3 siz
reen.
3. Image
Please
Normal
l Rights Res
e store)
“excellent” a
o enable the
ed under a
selection b
ser to switch
ons during
bles the vid
hange of del
e still pictur
zes, L, M, a
ge quality sc
(Media Pla
e select
l Good Excelle
served.
at the
e user
ll the
utton
h to a
video
deo to
livery
res of
and S.
creen
ayer)
ent
Copyright Ⓒ 2014 NTT DOCOMO, INC. All Rights Reserved.
- 36 -
6.3.2. Volume reduction (optimization) At its service launch, d anime store was using PNG24. After confirming that visual
degradation was limited with PNG8, the store has been using a tool mainly based on
PNG8 compression to perform optimization. By means of the tool, the data volume of
the TOP page has been compressed to one-third of its original volume.
6.3.3. Utilization of browser cache “d anime Store” sets the cache period of images to a relatively longer time so that the
cache is still effective when the user revisits after several days.
6.4. Notices for users
As mentioned in Section 6.1, the quality of experience of users can be improved by
providing video using the delivery scheme and the bitrate suitable to the specific
network environment available for each user. In addition, it is necessary to encourage
users to be conscious of the suitability of the delivery scheme and the bitrate to their
network environment in order to promote measures to improve the quality of experience.
Certain means such as a notice on a web site will be effective to make more users aware
of such measures. Figure 6-2 shows a notice given for that purpose at d anime store as
an example.
The mechanism explained in Section 2.2 has made regular video viewers sensitive to
data volume. Therefore, they are actively utilizing Wi-Fi and selecting image quality.
This is why d anime store provides information on Wi-Fi utilization in the notice shown
in Figure 6-2.
Copyright Ⓒ 2014 NTT DOCOMO, INC. All Rights Reserved.
- 37 -
Figure 6-2 A notice introducing measures for comfortable video viewing (d anime store)
Change of communication speeds for Xi
If you are subscribed to a “Packet Flat- rate Service for Xi,”, a 128-kbps cap is imposed on the your communication speed for the
rest of the month when the volume of data used in the current month exceeds a certain threshold (3GB, 7GB).
If the speed is reduced to 128 kbps, you may find it difficult to view video properly. Therefore, we propose certain measures to
enable you to avoid such communication speed change and enjoy your comfortable Anime Store experience.
--Use different lines at different times--
1. Utilize wireless LAN (Wi-Fi) at home.
You can download the episodes of anime you want to view all together via wireless LAN! You can view them without doing
anything special within 48 hours after the download. Even after 48 hours, what you need is only to confirm the license.
2. Utilize docomo Wi-Fi outdoors
docomo Wi-Fi is available at over 70,000 locations, including train stations, airports, coffee shops, fast food restaurants, family
restaurants, and convenience stores. Start now and use WiFi for free forever! You are recommended to apply now!
http://www.nttdocomo.co.jp/info/news_release/2012/08/22_00.html
3. Utilize “Normal” mode when using the mobile phone line (Xi、3G).
With the “Normal” mode, you can view approximately 180 episodes before the data volume reaches 7GB.
--Examples--
・First, you download 10 episodes or so at home or via docomo Wi-Fi.
・Within 48 hours after the download, you view them without doing anything special.
・After 48 hours, only you have to do is license authentication via the mobile phone.
・As you view more of the downloaded episodes, you download another 10 episodes.
・Only for the episodes you haven’t downloaded, you watch them in the “Normal” mode over Xi or 3G.
--Estimated data volume--
“Excellent” (data volume/episode:300MB, 3GB: 10 episodes, 7GB: 24 episodes)
“Good” (data volume/episode:100MB, 3GB: 31 episodes, 7GB: 72 episodes)
“Normal” (data volume/episode: 40MB、3GB:77 episodes, 7GB: 179 episodes)
(Note) The runtime of one episode: approx. 30 minutes. Assumptions: Normal=40MB, Good=100MB, Excellent=300MB.
--Packet Flat-rate Service for Xi--
For details, please visit DOCOMO web page shown below.
http://www.nttdocomo.co.jp/charge/packet/
--If you want to know your current data usage--
Data volume notice service
http://www.nttdocomo.co.jp/charge/online/notification_service/about/index.html
-->You will receive a notice when your data usage in the current month has reached[2GB, 3GB]or[6GB、7GB].
MyDocomo
http://www.nttdocomo.co.jp/charge/online/data_traffic/
-->By logging in with your docomo ID, you can see your accurate data usage status.
Data Volume Verification App
http://www.nttdocomo.co.jp/charge/online/data_app/
-->This app enables you to check an estimate of your packet data volume.
Copyright Ⓒ 2014 NTT DOCOMO, INC. All Rights Reserved.
- 38 -
6.5. Video specifications at “d anime store”
6.5.1. File specifications for HTTP streaming/Download technologies
“d anime store” is NTT DOCOMO’s video delivery service launched in July 2012.
This service is available for NTT DOCOMO’s Android terminals (DRM-capable models)
released in and after November, 2011. The video specifications for the Store have been
determined based on the specifications supported in playback on these terminal models
in accordance with the concepts shown in Table 6-2.
.
Table 6-2 Image quality concepts of d anime store
Image quality Concept
Normal This is the image quality that responds to the needs of users who want
to watch video even if its image quality is not entirely satisfactory
while on the move or in any other situations difficult to get stable
throughput in a LTE/3G environment.
Good This is the image quality that satisfies users who watch anime video
in a stationary position in a LTE/3G environment.
Excellent This is the image quality that provides users with the highest level of
satisfaction through a video-viewing terminal in a stable network
environment via WiFi, etc.
Based on the concepts above and the characteristics of anime content, d anime store
adopts parameter specifications different from those shown in 5.1.1. Specifically, the
Store uses different frame rates so that they can make the best use of limited bitrates to
realize “excellent images.”
Copyright Ⓒ 2014 NTT DOCOMO, INC. All Rights Reserved.
- 39 -
Table 6-3 Video specifications for H.264/AVC (d anime Store)
H.264/AVC, AAC-LC Image quality
Normal Good Excellent
Profile Baseline Baseline Baseline
Level 3 3 3
Resolution 16:9 428×240 428×240 720×404
4:3 320×240 320×240 640×480
Bitrate
(kbps)
Video 200 500 1,500
Audio 48 96 192
Total 248 596 1,692
Video Frame rate 20 20 29.97
Audio
Sampling frequency
(KHz) 44.1 44.1 48
Stereo/Monaural Monaural Stereo Stereo
As the models for 2014 summer or later support H.265/HEVC, d anime Store will
expand its content lineup supporting H.265/HEVC. The content specifications for
H.265/HEVC have been determined in light of the concepts above to reduce the amount
of data and achieve “high-speed download.” The same frame rates are used for all the
levels of normal, good, and excellent because verification results showed that higher
frame rates had little impact on the image quality. Based on the concepts above and the
characteristics of anime content, the Store adopts parameter specifications different
from those shown in 5.1.2.
The following table shows video specifications of d anime store for H.265/HEVC.
Copyright Ⓒ 2014 NTT DOCOMO, INC. All Rights Reserved.
- 40 -
Table 6-4 Video specifications for H.265/HEVC (d anime store)
H.265/HEVC, AAC-LC Image quality
Normal Good Excellent
Profile Main Main Main
Level 3 3 3.1
Tier Main Main Main
Resolution 16:9 428×240 428×240 852×480*
4:3 320×240 320×240 640×480
Bitrate
(kbps)
Video 100* 250* 700*
Audio 48 96 192
Total 148 346 892
Video Frame rate 29.97* 29.97* 29.97
Audio
Sampling frequency
(KHz) 44.1 44.1 48
Stereo/Monaural Monaural Stereo Stereo
*Different from H.264/AVC
6.5.2. File specifications for HLS technology
“d anime store” uses HLS for video streaming for iOS in accordance with the rules of
Apple Inc. The Store delivers video as a package of four files, which includes an
audio-only file according to the rules of Apple Inc. in addition to the three video files
shown in Table 6-3. The rules say that the size of an audio-only file “should never exceed
64 kbps even for a moment,” the Store encodes the audio data with an average rate of 32
kbps so that it will not exceed the maximum rate of 64kbps at any time. This is because
the file has been generated with VBR (Variable Bit Rate) due to such reasons as
encoding and software-related work, although generating a 64-kbps file with CBR
(Constant Bit Rate) may seem more reasonable.
The m3u8 file that integrates all the streams lists them in the order of excellent, good,
normal, and audio only. This arrangement has been made to accommodate customer
requests that they “want to enjoy anime with image quality as high as possible.”
Copyright Ⓒ 2014 NTT DOCOMO, INC. All Rights Reserved.
- 41 -
Figure 6-1 File structure for HLS (d anime store Excellent)
On the other hand, some customers were asking for a reduction in the volume of LTE
communication and a capability to specify service quality by themselves. NTT
DOCOMO altered the Store mechanism in April 2014 so that customers can set the
highest image quality level by themselves.
・ Excellent: (Excellent--> Good -- > Normal -- > Audio only)
・ Good : (Good -- > Normal -- > Audio only)
・ Normal : (Normal -- > Audio only)
6.6. 【Information】Introduction of tools
This section introduces some major tools that can be used to achieve the measures
described in Section 5.3. Please note that the links shown below may be subject to
change. Also, please be advised to use any of the tools in accordance with their terms of
use or licenses specified by their respective owners. NTT DOCOMO assumes no
responsibility for any damages, troubles, or other consequences resulting from the use
of the tools.
Still picture volume reduction tool:
modpagespeed: https://code.google.com/p/modpagespeed/
Smush.it: http://www.smushit.com/ysmush.it/
Web page improvement tools:
PageSpeed Insights: https://developers.google.com/speed/pagespeed/insights/
Yslow: http://developer.yahoo.com/yslow/
m3u8
動画( 32K)
※ 音声のみ
m3u8
動画( 普通)
m3u8
動画( )
m3u8
動画( すごくきれい)
m3u8
m3u8
Video (32k)
*Audio only
m3u8
(Normal)
Video
m3u8
Video
m3u8
Video
m3u8(Integrated)
(Excellent)(Good)
Copyright Ⓒ 2014 NTT DOCOMO, INC. All Rights Reserved.
- 42 -
Acknowledgements
NTT DOCOMO would like to extend our deepest appreciation to J-Stream Inc.,
DWANGO Co., Ltd., Forecast Communications Inc., and NTT Plala Inc. for their
valuable insights provided during the creation of this document.
NTT DOCOMO is also grateful to NTT’s Network Technology laboratories, Service
Evolution Laboratories, and Media Intelligence Laboratories for their technical
assistance provided for us in preparing this document.